the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 11 Dec 2023
The correlation between Arctic sea ice, cloud phase and radiation using A-train satellites
Abstract. Climate warming has a stronger impact on Arctic climate and sea ice cover (SIC) decline than previously thought. Better understanding and characterizing the relationship between sea ice, clouds and the implications for surface radiation is key to improving our confidence in Arctic climate projections. Here we analyze the relationship between sea ice, cloud phase and surface radiation over the Arctic, defined as north of 60° N, using active- and passive-sensor satellite observations from three different datasets. We find that all datasets agree on the climatology and seasonal variability of total and liquid-bearing (liquid and mixed-phase) cloud covers. Similarly, our results show a robust relationship between decreased SIC and increased liquid-bearing clouds in the lowest levels (below 3 km) for all seasons but summer, while increased SIC and ice clouds are positively correlated in two of the three datasets. A refined spatial correlation analysis indicates that the relationship between SIC and liquid-bearing clouds can change sign over the Bering, Barent and Laptev seas, likely because of intrusions of warm air from low latitudes during winter and spring. Finally, the increase of liquid clouds resulting from decreasing SIC is associated with enhanced radiative cooling at the surface, which should contribute to dampening future Arctic surface warming as SIC continues to decline.
- Preprint
(3015 KB) - Metadata XML
-
Supplement
(929 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2023-2940', Anonymous Referee #1, 05 Feb 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2023/egusphere-2023-2940/egusphere-2023-2940-RC1-supplement.pdf
-
CC1: 'Comment on egusphere-2023-2940', Luca Lelli, 08 Feb 2024
This commentary is not intended to be a comprehensive review but rather to comment on some aspects that, as a cloud researcher in the Arctic myself, I find interesting or worthy of improvement and explanation. My main points are three: the PHACT dataset, the error treatment and the conclusions drawn from the analysis of SIC and cloud radiative effect.
- The PHACT dataset
L43-44: "PHACT, which will be described in a separate paper".
It is peculiar to read the analysis of a dataset without having the ability to read the details of the algorithm used to create the same dataset. In my opinion, the authors could describe the algorithm in more detail, in the absence of a peer-reviewed article supporting the use of the dataset in this publication, to equip the reader with the necessary tools to understand the reach and the limitations of the dataset.
L45-46: "Cloud phase diagnostics are based on the cloud particle sphericity instead of temperature, in contrast with many passive sensors".
It is possible to differentiate the phases not only by means of temperature profiles but also by differential complex refractive index at those wavelengths where ice and water absorption differ. What accuracy is expected with the proposed approach compared to others? What ranges of sphericity values are assigned to categorize the variety of droplets and crystals and everything in between? - Errors
Within PHACT, the instantaneous profiles are aggregated at a spatial resolution of 2.5 degree. Recent studies (Kotarba, AR, 2022; Kotarba, AMT, 2022) have highlighted some shortcomings in the representation of climatological cloud fields by means of spaceborne lidars as function of grid resolution and cloud regimes. I think these results are relevant for the purposes of this article because it is clear that the errors introduced by undersampling (revisit and transect) can be considerable. I refer specifically to the figure 3 in the AMT article. The plots 3a-3c show the magnitude of cloud amount error with respect to general cloud regime (3a) and to sampling frequency (3c). I conclude that for high cloud amount CALIPSO sampling would result in absolute error of 5%, whereas for cloud amount of 60-80% the expected error would be 6-8% of cloud amount itself.
Two recent papers noted these results (Winker et al, ESSD, 2023; Bertrand et al, ESSD, 2023) and introduced correction factors to account for CALIPSO undersampling. It would be interesting to know the authors' opinion about this aspect and how this feature of CALIPSO impacts aggregate climatological cloud cover fields, differences with other datasets, and, finally, correlation values with sea ice extent.
This consideration brings me to ask what is the standard deviation of the curves in Figure 2 of this paper. The authors seem satisfied with the agreement among the datasets and it is certainly true that the seasonality is indeed in phase. But the absolute values are not in agreement and sometimes even exceeds the errors reported above in Kotarba AMT 2022. What if not even the standard deviation of the respective datasets overlap? - Conclusions
3-a) The sentence at line 182-183 ("Such a cooling effect is found in all seasons but winter, when the LW CRE warming exceeds the SW CRE cooling.") seems inaccurate to me.
Arctic winter months are characterized by the polar night. Assuming the authors calculate the net CRE at the surface from the difference between the downward and upward flux components of SW and LW for all-sky and clear-sky conditions (is this how you compute CRE?), the absence of sunlight makes the terms SW_down and SW_up zero and only the LW emissive terms are present. Cloud cannot reflect shortwave radiation during the polar night, right?
3-b) The last paragraph of the conclusion supports the results in Lelli et al. 2023.
A paper comes alive only if it ventures outside its own scientific project vacuum and it is related to previous research. Therefore, I think that the authors of this interesting manuscript could try to answer the following questions: (i) do your results contradict, confirm, strengthen previously established knowledge? (ii) which new research questions arise from the contrast between previous research and results here? (iii) How could further your research help resolve them? A better connection with past results is beneficial.
For instance, during the preparation of Lelli et al ACP 2023, we created the following figure where we look at CRE SW and LW above sea ice areas. The right panels show that the (negative) SW CRE trend increases for a decreasing sea ice extent, while the LW CRF trends slightly increase. Inspecting the crossing point between trends in SW and LW CRF as function of SIC, we see that already for a SIC change of 0.05%/month suffices for the SW CRF to offset the LW CRF. In these results the CRF changes as a function of cloud property changes are clearly conflated. Note that AMJ and JAS stand for Arctic spring (AMJ) and summer (JAS) and some differences might be due to the different season definition.
We additionally point the authors to Sections 3.2 onward in Lelli et al. 2023 for a detailed description of those cloud property changes and the CRE sensitivities. As a side note, cloud properties and fluxes have been extensively validated. The references of interest are given in Lelli et al. ACP 2023.
3-c) From what I can understand from the present manuscript, the authors analyzed CRE by relating it to the surface, treating it as if it were binary: no sea ice/sea ice. Where we know that in reality Arctic sea ice is a much more complex surface realm (leads, melt ponds, floes ... ) such that it likely results in distribution of ice concentration (0-100%) within a 2.5 degree side-by-side grid cells of the chosen aggregation. Would it be possible for the authors to add a more refined dimension of analysis and treat Arctic sea ice (and related trends) as a continuum? How would the result change?
3-d) I would like to ask whether the albedo of the analyzed sea ice surfaces can change the results found by the authors. CRE is by definition a net quantity, so it is to be related to the effectiveness of the surface to reflect SW. And this is a function of albedo, which in turn changes depending on whether - for example - there is fresh snow in the spring or old snow in the summer before the start of the melt season and so on. Can the authors comment on why they do not consider this aspect?
3-e) I think that is a bit of a missed opportunity - yet at authors' discretion - not to present also quantitative results on optical thickness of the clouds. In this way, the claims of this paper could be more substantiated and directly linkable to Lelli et al.
REFERENCES
Lelli, L., Vountas, M., Khosravi, N., and Burrows, J. P.: Satellite remote sensing of regional and seasonal Arctic cooling showing a multi-decadal trend towards brighter and more liquid clouds, Atmos. Chem. Phys., 23, 2579–2611, https://doi.org/10.5194/acp-23-2579-2023, 2023.
Bertrand, W., Kay, J. E., Haynes, J., & de Boer, G. (2023). A Global Gridded Dataset for Cloud Vertical Structure from Combined CloudSat and CALIPSO Observations. Earth System Science Data Discussions, 2023, 1-21.
Winker, D., Cai, X., Vaughan, M., Garnier, A., Magill, B., Avery, M., & Getzewich, B. (2023). A Level 3 Monthly Gridded Ice Cloud Dataset Derived from a Decade of CALIOP Measurements. Earth System Science Data Discussions, 2023, 1-37.
Kotarba, A. Z. (2022). Errors in global cloud climatology due to transect sampling with the CALIPSO satellite lidar mission. Atmospheric Research, 279, 106379.
Kotarba, A. Z.: Impact of the revisit frequency on cloud climatology for CALIPSO, EarthCARE, Aeolus, and ICESat-2 satellite lidar missions, Atmos. Meas. Tech., 15, 4307–4322, https://doi.org/10.5194/amt-15-4307-2022, 2022.
Citation: https://doi.org/10.5194/egusphere-2023-2940-CC1 - The PHACT dataset
-
RC2: 'Comment on egusphere-2023-2940', Anonymous Referee #2, 13 Feb 2024
Summary:
This study compares three different datasets of cloud fraction from satellites, including partitioning by phase in the Arctic. The responses of cloud fraction and cloud radiative effect to changes in sea ice concentration (SIC) are quantified. Much of the results are consistent with past work, e.g., the seasonality of correlations between SIC and liquid-containing clouds. Unique to this work is comparing results from active and passive sensors, as well as quantifying the response of ice clouds to SIC in the analysis.Main comments:
Other works have used quite specific methods to isolate the response of clouds to SIC changes and avoid confusing the cloud response to sea ice with the cloud response to atmospheric variability, e.g., the intermittent mask in Morrison et al. (2018) and the marginal ice zone crossing events in Taylor and Monroe (2023), along with instantaneous observations. First, this work uses some thresholding for grid cells based on SIC in Figure 7 and associated conclusions. Are the results, particularly for ice clouds, sensitive to this choice? Second, after (re)reading the Datasets section and data availability statement, it’s not actually clear to me what time resolutions the data used are. I assumed with using CALIPSO and CloudSat products the observations would be instantaneous, but the ERA5 SIC are monthly means. Could you please clarify this in the Datasets section?I was taught to be suspicious of cloud products from passive sensors in the Arctic, especially with changes in SIC, due to better detectability over open ocean compared to clouds (e.g., Liu et al. 2010). Is that not a concern with the CERES dataset used here, or could it impact any of the results?
Minor comments:
To some extent this is personal preference, but adding subplot labels and referencing them in the text might help make it easier for readers to connect the text to specific figures.Line 18: More recently Middlemas et al (2020) also found that cloud feedbacks have little impact on Arctic amplification.
Line 60: 20% seems like a substantial difference, no?
Line 85-6: Does "large agreement fraction with...phase retrievals" mean the two datasets agree well on the phase or also the amount of clouds?
Line 96: I don’t see a sharp contrast between land and ocean in DARDAR on the Pacific side of the Arctic. To me it looks like North America, Russia, and the ocean are all in the 70-80 CF bin?
Lines 106-8: Why the large difference between ground-based and space-based measurements? Is it a difference in location, land for ground-based but a mix of land and ocean for space-based?
Line 109: “show more”
Lines 112-3: Why does the CERES product have more ice cloud fraction than PHACT?
Lines 118-9: There are still differences in magnitude, and some slight differences in seasonality (e.g., some minima occur in March and April). Are you able to explain these differences?
Figure 2: I’m not sure if it would show a meaningful difference, but do the seasonal cycles from CERES and PHACT change is you compare the same years as DARDAR?
Lines 128-9: Shouldn’t the cloud response to SIC be regional/local? Is something else going on if the response is larger for Arctic-wide averages?
Line 140: In DARDAR winter and summer the maps don't look obviously positive or negative to me but more like noise.
Line 141: Do you have an explanation for the different responses of ice clouds to SIC in the different datasets?
Line 149: Are the results sensitive to the SIC thresholds used, specifically if you used more strict definitions of open ocean (e.g., SIC<15%) and ice-covered (e.g., SIC>80%)? Also, why these values? I might expect SIC near 0.4 and 0.6 to have mixed influences of ocean and sea ice.
Lines 153-4: I see that total CF changes look like liquid CF changes in PHACT (upper row), but to me it appears the DARDAR has more changes in ice (bottom row, second from the left)? Am I mis-interpreting the figure? Is there an explanation for largest changes in ice CF in DARDAR?
Lines 160-164: How much of the SW difference is changing clouds vs different surface albedo? Like you say, SW CRE can change just based on the surface albedo changing even if clouds remain the same.
Line 171: Did lines 111-3 mean that the larger ice CF in DARDAR is inaccurate?
Line 180: The correlation between ice clouds and SIC seems dependent on season and dataset – maybe add some qualifiers to this statement?
Line 184: Extra “and” between “declines and will”
References:
Liu, Yinghui, et al. "Errors in cloud detection over the Arctic using a satellite imager and implications for observing feedback mechanisms." Journal of Climate 23.7 (2010): 1894-1907.Middlemas, E. A., et al. "Quantifying the influence of cloud radiative feedbacks on Arctic surface warming using cloud locking in an Earth system model." Geophysical Research Letters 47.15 (2020): e2020GL089207.
Citation: https://doi.org/10.5194/egusphere-2023-2940-RC2 - AC1: 'Comment on egusphere-2023-2940', Gregory Cesana, 26 Mar 2024
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 402 | 89 | 26 | 517 | 36 | 15 | 13 |
- HTML: 402
- PDF: 89
- XML: 26
- Total: 517
- Supplement: 36
- BibTeX: 15
- EndNote: 13
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1